“Ripple effects” and forecasting home prices in Los Angeles, Las Vegas, and Phoenix

Abstract

We examine the time-series relationship between housing prices in Los Angeles, Las Vegas, and Phoenix. First, temporal Granger causality tests reveal that Los Angeles housing prices cause housing prices in Las Vegas (directly) and Phoenix (indirectly). In addition, Las Vegas housing prices cause housing prices in Phoenix. Los Angeles housing prices prove exogenous in a temporal sense and Phoenix housing prices do not cause prices in the other two markets. Second, we calculate out-of-sample forecasts in each market, using various vector autoregessive (VAR) and vector error-correction (VEC) models, as well as Bayesian, spatial, and causality versions of these models with various priors. Different specifications provide superior forecasts in the different cities. Finally, we consider the ability of theses time-series models to provide accurate out-of-sample predictions of turning points in housing prices that occurred in 2006:Q4. Recursive forecasts, where the sample is updated each quarter, provide reasonably good forecasts of turning points. Journal of Economic Literature Classification: C32, R31

The Time-Series Properties on Housing Prices: A Case Study of the Southern California Market

“…For the Bayesian models, we estimate Bayesian VAR (BVAR) and VEC (BVEC) models as well as BVAR and BVEC models that include spatial and causality priors 1 We exclude the El Centro MSA because of too short a time series on housing prices. (LeSage 2004, Gupta andMiller 2009). A causality BVEC model performs the best across all eight MSAs, although the forecasting performances in the individual MSAs do differ.…”

“…The discussion in this section relies heavily on LeSage (1999), Gupta and Sichei (2006), Gupta (2006), and Gupta and Miller (2009). VAR and VEC models typically use equal lag lengths for all variables in the model, which implies that the researcher must estimate many parameters, including many that prove statistically insignificant.…”

“…In sum, the 0.5 weights in the specification shown in equation (5) become 1.0 for neighbors and 0.1 for non-neighbors. Gupta and Miller (2009) …”

“…For the BVAR models, we start with a value of w = 0.1 and d = 1.0, and then increase the value to w = 0.2 to account for more influences from variables other than the first own lags of the dependant variables of the model. In addition, as in Dua and Ray (1995), Gupta and Sichei (2006), Gupta (2006), and Gupta and Miller (2009), we also estimate a BVAR model with w = 0.3 and d = 0.5. We also introduce d = 2 to increase the tightness on lag m. Finally, we specify σ c =0.1, η=8, θ=0.5 for the random-walk models with the two different specifications for causality and spatial priors.…”

Heterogeneous convergence of regional house prices and the complexity in China

“…The disadvantage of these methods is about the DOF constraints for these models can't contain too much regional units (typically less than 8). The third category is spatial or temporal models (Gupta and Miller, 2012). These models have to be set based on spatial weights matrix, which is generally defined by "geographical adjacent" or "economic proximity".…”